Use Calibration Training to Catch a Bad Values-Spending Guess

What each one is actually for

Values-Based Spending

Values-Based Spending works as a single pause inserted between wanting something and paying for it: before you complete a purchase, you name the value you believe it serves — connection, health, creative work, comfort — from a short list you wrote down in advance, not invented on the spot to justify what you already want. If nothing on the list fits, or the fit feels thin once you say it out loud, that is the signal to wait rather than buy. The mechanism is friction on purpose: a deliberate half-minute of naming inserted where impulse would otherwise run straight to checkout, aimed at redirecting spending toward purchases that hold up days later rather than ones that felt good only for the ten minutes around clicking buy.

Superforecasting

Calibration training is the specific practice behind the research on 'superforecasters,' people whose predictions about world events consistently outperform experts and pundits. Instead of saying 'I think this will probably happen,' you commit to a number — 'I am seventy percent confident this will happen by March' — and you write it down where you can check it later. Over dozens of predictions, you compare your stated confidence levels against how often you were actually right: if the things you called seventy percent confident only came true forty percent of the time, you are overconfident, and the training's whole point is to notice that gap and correct it, rather than simply trusting your gut confidence at face value forever.

What you get out of pairing them

A values-check before spending is a prediction dressed up as a principle: you are claiming that this purchase, of the things on your list, will actually deliver what it promises. Almost nobody ever goes back and checks whether that specific prediction was right. You buy the thing 'for connection' or 'for health,' and then life moves on, and the claim just sits there unverified, available to be repeated with just as much unearned confidence the next time. Calibration training gives you the missing half: a habit of writing the prediction down with a number attached, then actually checking weeks later whether it held up. Do that across enough purchases and you stop having a vague, static sense of 'I'm someone who spends on what matters' and start having an actual, improving track record — the specific categories where your predictions run reliably strong, and the specific categories, often the ones tied to a mood rather than a plan, where your ninety-percent-confident guesses turn out right closer to half the time.

What Values-Based Spending rests on

• Self distancing improves judgment randomised trials
  Evaluating a situation from a distanced, third-person vantage point instead of an immersed first-person one improves the quality of judgment about it.
• Friction before purchase reduces impulse spending plausible mechanism
  Inserting a deliberate values-check step between impulse and purchase adds friction that reduces low-value impulsive spending.
• Values alignment check increases satisfaction observational studies
  Checking a planned purchase against stated personal values before committing redirects spending toward choices that produce more lasting satisfaction.

What Superforecasting rests on

• Decomposition improves estimate accuracy observational studies
  Breaking a complex uncertain question into smaller more tractable sub-questions produces more accurate estimates than judging the whole question at once.
• Explicit probability estimates improve calibration randomised trials
  Forcing a numeric probability rather than a vague verbal hedge makes a prediction falsifiable and trainable.
• Outcome feedback corrects judgment bias randomised trials
  Regularly scoring predictions against actual outcomes and reviewing the misses corrects systematic biases in judgment over time.

Where they actually meet

Both practices are trying to pull a decision out of the immersed, first-person moment where it is easiest to fool yourself and into a more distanced vantage point where it can actually be judged. Values-based spending does this by making you name the value out loud before you buy, rather than letting the purchase happen inside the felt urgency of wanting it. Calibration training does the identical thing to a belief: instead of trusting a gut sense of confidence from inside it, you step back, assign it a number, and let that number be checked later by someone who is, in effect, a more distant version of yourself looking back. Naming a value and naming a probability are the same move on different material — both take a claim you would otherwise just feel and hold it far enough away to examine.

The principle underneath both

A claim you only ever feel, whether it is 'this purchase matters to me' or 'this is likely to happen,' is protected from correction for exactly as long as it stays a feeling. The moment it becomes a specific, stated claim — a named value, a named number — it can be checked against what actually happens, and being checkable is what lets a person's judgment about their own future actually improve over time rather than staying whatever it happened to be on day one.

What that makes the real goal

If turning a felt claim into a checkable one is what does the work, the objective both practices are reaching for is the same: a track record of your own predictions about what will satisfy or hold up, specific and numbered enough that being wrong actually teaches you something, rather than a permanent, untested sense that you are generally a reasonable judge of your own future. Values-based spending needs this to stop being a one-time act of naming and start being a habit that gets sharper. Calibration training needs a real domain to apply itself to beyond abstract world events, and personal spending decisions, repeated weekly, supply exactly that.

Other practices that serve the same goal

Prediction markets and betting pools pursue the same objective at a larger scale, forcing a crowd of people to put a number on an uncertain outcome and then paying out based on what actually happened, which is one of the most consistently well-evidenced ways to produce accurate collective forecasts, though it is built for group questions rather than a single person's private spending choices. Post-purchase satisfaction tracking, sometimes used in behavioral economics research, asks people to predict how much they will enjoy a purchase and then rate it again after a set number of weeks, which targets the same values-and-money gap directly and has its own separate research base showing people are often surprised by their own results. Habit-tracking apps that log a stated intention against what actually happened over time apply a lighter, less numeric version of the same feedback loop to everyday behavior generally, not spending specifically, and mostly rely on user-reported outcomes rather than controlled trials.

Using values-based spending to give calibration training something concrete to grade

Calibration training usually practices on abstract world events — elections, sports outcomes, geopolitical questions — which is useful for the skill but gives you no personal stake and no fast feedback; you might wait months to learn whether your seventy-percent call was any good. A values-check before a purchase is a prediction you can grade within weeks. Before you buy, write the specific claim down with a number: 'I am eighty percent confident this will still feel worth it in a month, because it serves the value I named.' Set a reminder for four weeks out, and when it fires, write down what actually happened — not just whether you liked the purchase, but whether your stated confidence was the right confidence. Do this for ten or fifteen purchases and you will likely find your calibration is uneven across categories: reliably accurate about purchases tied to a concrete plan, and far less accurate about the ones bought while stressed or scrolling late at night, where confidence was high and follow-through on satisfaction was not.

Using calibration training to sharpen the values-check itself

One habit borrowed directly from calibration training strengthens the values-check before it even happens: decomposition, meaning breaking one big uncertain question into smaller ones that are each easier to judge accurately than the whole question at once. Instead of asking the single vague question 'does this serve my values,' break it into three narrower ones — will I still want this in a week, does it require anything from me afterward that I am actually likely to follow through on, and is there a cheaper or free version of the same value I have not tried first. Answering three specific, smaller questions produces a more honest read than one big feeling ever will, because a vague global question is exactly the kind of question forecasting research shows people answer worst — and it is where a values-check quietly turns into a justification for a purchase you had already decided to make.

A cart, a countdown timer, and a number you have to write down

You are looking at a course bundle on sale, a countdown timer ticking down in the corner of the page, and the value you would name if asked is 'growth' — it is on your list, so the purchase clears the values-check almost instantly. Before you check out, you make yourself write the actual claim with a number attached: 'I am ninety percent confident I will have started this within two weeks.' Writing the number instead of just feeling confident makes you notice something the feeling alone skipped past — you have bought three similar bundles in the last year and started exactly one of them. The ninety percent was borrowed from how excited you feel right now, not from your own track record, which the calibration habit exists specifically to catch. You lower the number to what your actual history supports, closer to thirty percent, and at thirty percent the purchase does not clear the bar the way it did at ninety. You close the tab, not because growth stopped mattering, but because your own numbers, not your mood, just told you the truth about how this one was likely to go.

When this is worth reaching for

The same category keeps recurring — a course, gear for a hobby restarted more than once, a subscription justified by the same value every time — and each purchase feels new even though the pattern isn't; a written, numbered prediction catches what a fresh in-the-moment feeling never will. It only pays for itself once the purchase is large or recurring enough to be worth the extra thirty seconds; for genuinely small, low-stakes buys the overhead costs more than it returns.

Where this does not apply

For routine, low-value purchases — groceries, a coffee, a bus fare — running a full values-check with a numeric confidence estimate attached is more process than the decision deserves, and turning every small purchase into a graded prediction is a fast way to make the whole habit feel exhausting enough to quit. This pairing also is not a fit for someone using precise numbers to dress up a decision they have already emotionally made; if the number always seems to land wherever supports the purchase you wanted anyway, the calibration step has stopped doing its job and is worth dropping until you can write the number down honestly, before you know which answer you want.

What the evidence does and does not support

Calibration training's core finding, that tracking numeric predictions against outcomes measurably improves judgment over time, comes from a substantial research program, most visibly the Good Judgment Project, and rests on real randomized comparisons between trained and untrained forecasters. Values-based spending's satisfaction claims mostly rest on behavioral-economics observation about the mismatch between predicted and experienced satisfaction from purchases, rather than on trials of the specific practice of naming a value before buying. Nobody has run a study that hands a personal values-check a calibration score and measures whether spending choices actually improve as a result — this page is describing a plausible personal experiment built from two separately evidenced literatures, not a tested combination.